llvm dev - Aug 2014 - [LLVMdev] Matching up inlined basic blocks with original basic blocks.

Hello All,

I have some data tied to the basic blocks in a function, and after inlining
that function, I'd like to recover that data in the inlined version.  Is
there some way to match up the inlined version of the function with the
original basic blocks?

Thanks,
Jeremy
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140803/bb5c608f/attachment.html>

On Sun, Aug 3, 2014 at 6:36 PM, Jeremy Salwen <jeremysalwen at gmail.com>
wrote:
> Hello All,
>
> I have some data tied to the basic blocks in a function, and after
> inlining that function, I'd like to recover that data in the inlined
> version.  Is there some way to match up the inlined version of the function
> with the original basic blocks?
>
See the code in CloneFunctionInto and CloneAndPruneFunctionInto. This is
what does the actual inlining. You can see how that code tracks the basic
blocks after inlining.

Honestly, the right way to do this architecturally is for us to support
metadata attached to basic blocks (if we don't already?) and then for the
inliner to preserve it when inlining. Then you could use metadata as a
container for your data.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140803/aa804fe8/attachment.html>

----- Original Message -----
> From: "Chandler Carruth" <chandlerc at google.com>
> To: "Jeremy Salwen" <jeremysalwen at gmail.com>
> Cc: "llvmdev" <llvmdev at cs.uiuc.edu>
> Sent: Sunday, August 3, 2014 9:18:38 PM
> Subject: Re: [LLVMdev] Matching up inlined basic blocks with original basic
blocks.
> 
> On Sun, Aug 3, 2014 at 6:36 PM, Jeremy Salwen <
> jeremysalwen at gmail.com > wrote:
> 
> 
> 
> Hello All,
> 
> I have some data tied to the basic blocks in a function, and after
> inlining that function, I'd like to recover that data in the inlined
> version. Is there some way to match up the inlined version of the
> function with the original basic blocks?
> See the code in CloneFunctionInto and CloneAndPruneFunctionInto. This
> is what does the actual inlining. You can see how that code tracks
> the basic blocks after inlining.
> 
> 
> Honestly, the right way to do this architecturally is for us to
> support metadata attached to basic blocks (if we don't already?) and
> then for the inliner to preserve it when inlining. Then you could
> use metadata as a container for your data.
We don't currently support metadata on basic blocks. There was a patch for
this some time ago,
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20120507/142379.html.

 -Hal
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> 
-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory

Hi Chandler,

It seems that both functions use a VMap argument in which they create a map
from old values to their cloned copies.  This would work great, however,
when I look in InlineFunction() (InlineFunction.cpp line 794), the VMap
seems to be hidden from the caller, and there is an explicit comment "Scope
to destroy VMap after cloning".

 Is InlineFunction() the wrong entry point for inlining a function?  Or do
I need to create a modified version that exposes the VMap?

Thanks,
Jeremy





On Sun, Aug 3, 2014 at 10:18 PM, Chandler Carruth <chandlerc at
google.com>
wrote:
>
> On Sun, Aug 3, 2014 at 6:36 PM, Jeremy Salwen <jeremysalwen at
gmail.com>
> wrote:
>
>> Hello All,
>>
>> I have some data tied to the basic blocks in a function, and after
>> inlining that function, I'd like to recover that data in the
inlined
>> version.  Is there some way to match up the inlined version of the
function
>> with the original basic blocks?
>>
>
> See the code in CloneFunctionInto and CloneAndPruneFunctionInto. This is
> what does the actual inlining. You can see how that code tracks the basic
> blocks after inlining.
>
> Honestly, the right way to do this architecturally is for us to support
> metadata attached to basic blocks (if we don't already?) and then for
the
> inliner to preserve it when inlining. Then you could use metadata as a
> container for your data.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140804/472887bf/attachment.html>